Body-Image and Eating Disturbances Predict Onset of Depression ...

Page 1 of 12

Journal of Abnormal Psychology August 2000 Vol. 109, No. 3, 438-444

© by the American Psychological Association For personal use only--not for distribution.

Body-Image and Eating Disturbances Predict Onset of Depression Among Female Adolescents A Longitudinal Study Eric Stice Department of Psychology University of Texas at Austin Chris Hayward Department of Psychiatry Stanford University School of Medicine Rebecca P. Cameron Department of Psychology University of San Francisco Joel D. Killen Department of Medicine Stanford University School of Medicine C. Barr Taylor Department of Psychiatry Stanford University School of Medicine ABSTRACT This study examined data from a 4-year school-based longitudinal study ( n = 1,124), to test whether the increase in major depression that occurs among girls during adolescence may be partially explained by the body-image and eating disturbances that emerge after puberty. Elevated body dissatisfaction, dietary restraint, and bulimic symptoms at study entry predicted onset of subsequent depression among initially nondepressed youth in bivariate analyses controlling for initial depressive symptoms. Although the unique effect for body dissatisfaction was not significant in the multivariate model, this set of risk factors was able to fairly accurately foretell which girls would go on to develop major depression. Results were consistent with the assertion that the body-image- and eating-related risk factors that emerge after puberty might contribute to the elevated rates of depression for adolescent girls.

Major depression is one the most prevalent psychiatric disorders, with research suggesting that approximately 20% of adolescents meet lifetime criteria ( Lewinsohn, Hops, Roberts, Seeley, & Andrews, 1993 ). Additionally, adolescent depression is associated with suicide attempts and high rates of comorbid anxiety disorders, disruptive behavior disorders, and substance abuse and predicts future adjustment problems, including academic failure, marital difficulties, unemployment, substance abuse, delinquency, and legal problems ( Birmaher et al., 1996 ; Gotlib, Lewsinsohn, & Seeley, 1998 ; Newcomb & Bentler, 1988 ). The rate of depression increases dramatically for girls during the second decade of life, and by late adolescence, twice as many girls are depressed as boys ( Hankin et al., 1998 ; Lewinsohn et al., 1993 ). However, the explanation for the precipitous rise in depression for girls during adolescence has been elusive ( Nolen-Hoeksema & Girgus, 1994 ). One possibility is that females face additional risk factors for depression, above and beyond the ones they share with their male counterparts (e.g., negative life events and family history) and that these extra risk factors escalate in early adolescence. Such a gender-additive model may prove conceptually useful in explaining the dramatic rise in depression among girls. As has been suggested by others (e.g., NolenHoeksema, 1994 ), we posit that because puberty moves girls away from the current thin ideal, it

http://spider.apa.org/ftdocs/abn/2000/august/abn1093438.html

11/7/2001

Page 2 of 12

precipitates body-image- and eating-related risk factors that may contribute to the higher rates of major depression observed among girls during adolescence. In contrast, puberty moves boys toward the culturally prescribed ideal body for males and may not precipitate such additional risk factors for depression. It has been proposed in the dual pathway model of bulimic pathology ( Stice, Nemeroff, & Shaw, 1996 ) that elevated body mass, body dissatisfaction, and dieting contribute to affective problems for adolescent girls (which in turn promote the development of bulimic behaviors). Theoretically, elevated body mass results in body dissatisfaction because being overweight is not currently considered socially desirable. This body dissatisfaction in turn may contribute directly to depression, because appearance is a central evaluative dimension for females in Western cultures. Moreover, body dissatisfaction is also thought to foster dieting, which may in turn increase the chances of depression. Affective distress might result from the failures that can be associated with dietary efforts. Thus, body dissatisfaction is thought to mediate the relation of body mass to dieting and depression, and dieting is thought to partially mediate the relation of body dissatisfaction to depression. Accordingly, elevated body mass, body dissatisfaction, and dieting might be expected to predict the onset of depression among girls during adolescence. In addition, bulimic pathology, which also theoretically results from this set of risk factors, may contribute to the increased incidence of depression among girls during adolescence. For example, the shame and guilt that result from engaging in bulimic behavior might increase the risk for onset of major depression. Accordingly, this study investigates a set of body-image- and eatingdisturbance-related risk factors that may contribute to the high rates of depression for adolescent girls. Several lines of evidence are consistent with this explanation for the dramatic increase in depression for girls during adolescence. First, whereas girls generally dislike the physical changes that accompany puberty (e.g., the increase in adipose tissue), boys typically like the changes that come with puberty (e.g., the increase in muscle mass; Keel, Fulkerson, & Leon, 1997 ; Rierdan, Koff, & Stubbs, 1989 ). Second, depressive symptoms are correlated positively with body mass, body dissatisfaction, dieting, and bulimic pathology among samples of adolescent girls (e.g., Fabian & Thompson, 1989 ; Graber, Brooks-Gunn, Paikoff, & Warren, 1994 ; Killen et al., 1987 ; Rosen, Tacy, & Howell, 1990 ; Stice, Killen, Hayward, & Taylor, 1998b ). Third, body dissatisfaction prospectively has predicted depressive symptoms during adolescence for girls ( Allgood-Merten, Lewinsohn, & Hops, 1990 ; Cole, Martin, Peeke, Seroczynski, & Hoffman, 1998 ; Rierdan et al., 1989 ). Thus, there is some evidence that body-image and eating disturbances may be useful in explaining the escalation in depression observed among adolescent girls. However, much of this research is crosssectional and, consequently, cannot be unambiguously interpreted. For example, body dissatisfaction may be a consequence rather than a cause of depression. In addition, little research has examined the relation of a broad spectrum of body-image and eating-related factors, including dieting and bulimic pathology, to depression. Finally, much of the prior literature has focused on depressive symptoms rather than diagnostic levels of major depression (for a review, see Nolen-Hoeksema & Girgus, 1994 ). Although these past studies have contributed to our understanding of the correlates of depressive symptoms, it would be useful to ascertain whether these body-image and eating-related risk factors predict clinically significant levels of depression. Accordingly, we tested whether elevations in body mass, body dissatisfaction, dieting, and bulimic symptoms prospectively predicted onset of major depression among initially nondepressed adolescent girls. 1

Method Participants Participants were 1,124 female students from three northern California high schools, who ranged in age from 13.0 to 16.9 years ( M = 14.7) at study entry. Only female participants were included in this


11/7/2001

Page 3 of 12

investigation because most of the scales examined here were not administered to the male participants (e.g., the Restraint Scale) who participated in the longer study. The sample comprised 25% Asians, 4% Blacks, 42% Whites, 15% Hispanics, 7% Native Americans, 6% who specified mixed racial heritage, and 1% who specified "other." Maximum parental education ranged from less than high school (5%) to graduate degree (29%), which was also the mode. Procedures The study was presented to participants as an investigation of adolescent health beliefs and behaviors. A passive parental consent procedure was used, resulting in a participation rate of 95% of eligible students. Staff trained by the principal investigators conducted Time 1 and three annual follow-up assessments. At each assessment, students completed a self-report questionnaire, had their weight and height measured by female research assistants, and participated in a structured clinical interview during regular class times. Participants were identified by a special identification number to ensure confidentiality. There were varying lengths of follow-up because participants could enter the study at any point during the 4year study period and because of attrition. We attempted to follow students who dropped out of school and made special arrangements to assess students who had enrolled in continuation schools. This project was approved by the Stanford University Committee for the Protection of Human Subjects. Measures Body mass. We used body mass index (BMI = kg/M 2 ) to reflect adiposity ( Garrow & Webster, 1985 ). Height was measured to the nearest millimeter using a portable direct reading stadiometer. Students were measured without shoes and with their bodies positioned so that their heels and buttocks were against the vertical support of the stadiometer and their heads aligned so that the auditory canal and the lower rim of the orbit of the eye socket were in a horizontal plane. Body weight was assessed to the nearest 0.1 kg using digital scales, with the participants wearing light indoor clothing without shoes or coats. Two measures of height and weight were obtained and averaged for analyses. Research has documented that the BMI is a valid measure of adiposity ( Garrow & Webster, 1985 ; Kraemer, Berkowitz, & Hammer, 1990 ). Body dissatisfaction. We assessed body dissatisfaction with the 9-item Body Dissatisfaction subscale from the Eating Disorders Inventory (EDI; Garner, Olmsted, & Polivy, 1983 ). Items such as "I think my thighs are too large" are rated on a 6-point scale, ranging from 1 ( never ) to 6 ( always ), and were averaged for analyses. This scale possesses adequate reliability and validity ( Garner et al., 1983 ). Cronbach's alpha for this scale was .90 at Time 1. Dietary restraint. We used the Restraint Scale to assess dietary restraint ( Heatherton, Herman, Polivy, King, & McGree, 1988 ). A sample item from this scale is "How often do you diet?" with responses ranging from 1 ( never ) to 5 ( always ). The reliability and validity of this scale have been established ( Heatherton et al., 1988 ). Items were averaged for analyses. Cronbach's alpha for this scale was .79 at Time 1. Bulimic symptoms. We used the Bulimia subscale from the EDI to assess bulimic symptoms ( Garner et al., 1983 ). Items such as "I have gone on eating binges where I felt I could not stop" are rated on a 6-point scale, ranging from 1 ( never ) to 6 ( always ) and were averaged for analyses. This scale possesses acceptable


11/7/2001

Page 4 of 12

reliability and validity ( Garner et al., 1983 ). Cronbach's alpha for this scale was .81 at Time 1. Depressive symptoms. We used the Center for Epidemiologic Studies-Depression Scale (CES—D; Radloff, 1977 ) to assess depressive symptoms. Items such as "I have been feeling pretty down and unhappy this week" are rated on a 4-point scale, ranging from 0 ( never ) to 3 ( most of the time ) and were summed for analyses. The reliability and validity of this scale have been documented ( Robert, Lewinsohn, & Seeley, 1991 ). Cronbach's alpha for this scale was .91 at Time 1. Major depression diagnoses. We used the nonpatient version of the Structured Clinical Interview for DSM—III—R (SCID; Spitzer, Williams, Gibbon, & First, 1990 ) to assess major depression diagnoses. Female graduate-level interviewers underwent extensive training and were provided with regular supervision. The interview consisted of 12 questions assessing Diagnostic and Statistical Manual for Mental Disorders ( DSM— III—R; American Psychiatric Association, 1987 ) criteria for major depression. The Time 1 interview assessed lifetime criteria for major depression, whereas the Time 2 through Time 4 interviews assessed criteria over the past year only. This interview possesses acceptable reliability and validity ( Segal, Hersen, & Van Hasselt, 1994 ; Strakowski et al., 1993 ). We examined the test—retest reliability of diagnoses in this sample by reinterviewing a randomly selected subsample of 420 students approximately 3 days after their original interview. The kappa coefficient, which represents a chancecorrected level of agreement between the two interviews, was .51, which represents fair agreement according to the criteria proposed by Fleiss (1981) .

Results Preliminary Analyses Preliminary analyses tested for attrition biases that might have compromised the generalizability of the findings. The average annual attrition rate was 15% (see Hayward, Killen, Kraemer, & Taylor, 1998 , for greater detail), resulting in a total attrition rate of 39% over the 4-year study. The 15% average annual attrition rate primarily resulted because students moved (10%), but the remainder was due to absenteeism (3%) and refusal to participate (2%). One-way analysis of variance (ANOVA) models and chi-square analyses indicated that participants who failed to provide data at one or more of the assessments did not differ significantly from those who provided complete data on age, ethnicity, depressive symptoms, body mass, body dissatisfaction, dietary restraint, bulimic symptoms, or major depression diagnoses at study entry. However, there was evidence that participants who dropped from the study had a significantly lower level of parental education than those who did not drop, F (1, 1075) = 9.67, p < .01, although this effect accounted for only 2% of the variance. Thus, some caution should be exercised in generalizing the findings. Prospective Analyses Of the 1,124 female participants, 100 met criteria for a lifetime diagnosis of major depression at study entry (including past and current). These participants were excluded from analyses to ensure a prospective test of hypotheses. Of the remaining 1,024 participants, 80 were diagnosed with an onset of major depression during the study period. Table 1 contains the means and standard deviations for each of the putative risk factors and initial depressive symptoms, as well as the correlations among them.


11/7/2001

Page 5 of 12

The risk factors for onset of major depression were analyzed with Cox proportional hazards regression analyses ( Cox, 1970 ). This analytic technique, which is also known as survival analysis, has the advantage of allowing for varying lengths of follow-up in longitudinal studies and thus minimizes biases due to attrition and accommodates right censored cases that do not experience the target event during the study period ( Willett & Singer, 1993 ). Unlike most analytic techniques that require complete data for all time points (e.g., repeated measures ANOVAs), hazard analysis uses all available data at each time point because it treats missing data as right censored. Specifically, it calculates the hazard for onset only among asymptomatic participants who provided data at each particular time point. Hazard analysis produces unbiased estimates of survival time as long as the assumption that censoring is unrelated to the likelihood of the event (major depression diagnosis) is met ( Luke, 1993 ). This assumption appeared to be satisfied in this data set, as a chi-square analysis suggested that attrition was not significantly related to the likelihood of a major depression diagnosis during the years that participants provided data, χ 2 (1, N = 1124) = 0.73, ns. We first included each putative risk factor in an individual hazard model to provide a picture of the relations between each factor and the onset of major depression that was not complicated by multicollinearity. Again, all participants were initially free of major depression, and all putative risk factors were assessed at study entry. We also controlled for depressive symptoms at study entry in all analyses to ensure that the putative risk factors predict onset of major depression above and beyond the effects of initial level of depressive symptoms. Because preliminary bivariate hazard models indicated that adolescent ethnicity and parental education (a proxy for socioeconomic status) were not significantly related to onset of major depression, these demographic factors were not included as covariates. The top half of Table 2 contains the unstandardized parameter estimates, hazard ratios, 95% confidence intervals, and significance levels for the bivariate relations between risk factors and the probability, or hazard, for onset of major depression over the study period. As would be expected, initial elevations in depressive symptoms prospectively predicted a greater hazard for the subsequent onset of major depression. In keeping with hypotheses, initial elevations in body dissatisfaction, dietary restraint, and bulimic symptoms, but not body mass, at study entry each prospectively predicted a greater hazard for the subsequent onset of major depression when we controlled for the effects of initial depressive symptoms. 2 The hazard ratio reflects the magnitude of the relation between the risk factor and the hazard for onset of major depression. If 1 is subtracted from the hazard ratio and the result multiplied by 100, this number reflects the estimated percentage change in the hazard for onset of major depression for each unit increase in the risk factor. For example, 1.31 − 1 × 100 = 31%, indicating that the hazard for onset of major depression increases 31% for each unit increase on the Body Dissatisfaction scale. Although these estimates depend on the scale of the independent variable, it provides some indication of the magnitude of the effects. 3 Second, to examine the unique relations of the risk factors within a multivariate model, the three variables that showed significant effects were simultaneously included in a Cox proportional hazards regression analysis predicting onset of major depression. As before, all participants were initially free of major depression, we controlled for initial depressive symptoms, and all putative risk factors were assessed at study entry. The lower half of Table 2 contains the unstandardized parameter estimates, hazard ratios, 95% confidence intervals, and significance levels for the multivariate relations between risk factors and the hazard for onset of major depression over the study period. In the multivariate model, initial elevations in dietary restraint and bulimic symptoms showed significant relations to the onset of subsequent major depression. However, the effects for initial depressive symptoms and body dissatisfaction were no longer significant in the multivariate models.


11/7/2001

Page 6 of 12

Finally, to determine how well this set of body-image- and eating-related risk factors collectively discriminated between the initially nondepressed adolescents who went on to develop major depression from those who did not, we conducted a discriminant function analysis wherein body dissatisfaction, dietary restraint, and bulimic symptoms were used as predictor variables. This analytic technique determines which linear combination of factors best differentiates two or more groups ( Tabachnick & Fidell, 1989 ) and takes into account the intercorrelations among the independent variables. The program converged after one discriminant function was extracted (the maximum possible when there are only two groups), resulting in χ 2 (3, N = 1122) = 24.76, p < .0001. An examination of the structure matrix, or pooled within-groups correlations between the independent variables and discriminant function, revealed that body dissatisfaction ( r = .62), dietary restraint ( r = .88), and bulimic symptoms ( r = .76) all loaded significantly on the discriminant function. 4 Overall, this discriminant function showed moderate predictive accuracy in that it was able to correctly classify 66% of the adolescents into those who subsequently experienced onset of major depression and those who did not (compared with a 50:50 correct classification rate based on chance alone). Following Tatsuoka (1971) , we specified that the discriminant function analysis use a 50:50 prior probability for group membership, rather than the actual 92:8 base rate, because when the base rate of one group is extremely low, the program simply assigns all participants to the larger group to maximize predictive accuracy. We selected this approach because it did not seem reasonable to capitalize on this factor to improve our overall classification rate.

Discussion This study tested whether the increases in major depression among adolescent girls could be partially explained by the body-image- and eating-disturbance-related risk factors that emerge after puberty. In keeping with hypotheses, initial body dissatisfaction, dietary restraint, and bulimic symptoms, but not body mass, predicted onset of subsequent depression among initially nondepressed adolescents over the study period. Although one of the unique effects was not significant in the multivariate model because these risk factors were correlated, this set of factors was able to foretell which girls would go on to develop major depression with fair accuracy. The longitudinal design, wherein risk factors measured at study entry were used to predict onset of depression over the study period, while we controlled for initial depressive symptoms, permits stronger inferences about the direction of effects. Moreover, because the predictors were assessed through questionnaire, whereas depression was measured with structured psychiatric interviews, the possibility that method variance explains the findings is reduced. Nonetheless, it is not possible to rule out third-variable explanations that could account for the relations (e.g., genetic effects). The evidence that body dissatisfaction predicts onset of major depression in the bivariate analyses converges with past research indicating that body dissatisfaction prospectively predicts depressive symptoms during adolescence (e.g., Rierdan et al., 1989 ). Body dissatisfaction is thought to contribute to depression because appearance is a critical dimension on which females are evaluated in Western culture. Body dissatisfaction may also contribute to dietary restraint, which in turn is postulated to lead to increased depression because of the failures that are often associated with dieting efforts. Consistent with this last assertion, our findings also provide evidence that dietary restraint predicts onset of subsequent depression. To our knowledge, this is the first study that has generated prospective evidence of these relations using data from a community study of adolescents. Finally, results indicate that bulimic symptoms predict onset of depression. Theoretically, bulimic pathology contributes to depression because of the shame and guilt that are often associated with binge eating and purging. Again, this is a novel finding in that we were unable to locate prior studies that prospectively investigated this relation. It is noteworthy that initial dietary restraint and bulimic symptoms predicted onset of major depression in the multivariate analyses but that initial depressive symptoms did not, because studies have found that past depressive symptoms are typically the strongest risk factor for


11/7/2001

Page 7 of 12

onset of major depression (e.g., Lewinsohn et al., 1994 ). There are at least two possible interpretations for the finding that body dissatisfaction predicts onset of major depression in the bivariate analyses but not in the multivariate model. First, the effects of some of these risk factors may be mediated by the other risk factors. Indeed, the finding that the effect for body dissatisfaction became nonsignificant in the multivariate analysis is consistent with the mediational relations articulated in the dual pathway model of bulimia ( Stice et al., 1996 ). Specifically, this model asserts that the effects of body dissatisfaction on depression are partially mediated by increased dietary restraint. Thus, one would expect that the effects of these upstream variables would diminish or become nonsignificant when the downstream variables are entered simultaneously (see Baron & Kenny, 1986 ). Second, it may simply be that these risk factors co-occur (as reflected in Table 1 ), and this makes it difficult to detect unique effects of each risk factor when they are considered simultaneously. It would be useful for future research to investigate these two explanations for the relations among these risk factors. Contrary to expectations, elevated body mass did not predict onset of depression in the bivariate or multivariate analyses. This null finding, taken in conjunction with the positive effects for body dissatisfaction, suggest that cognitive aspects of body-image disturbance, rather than actual body dimensions, may play a more important role in the etiology of depression by midadolescence. It might also be the case that non-weight-related aspects of body-image disturbances (e.g., dissatisfaction with the appearance of certain body parts) are more important than weight per se in promoting depression. This interpretation is supported by the fact that body mass only accounted for 18% of the variance in body dissatisfaction. The evidence that bulimic symptoms predicted onset of depression, in conjunction with the finding that negative affectivity predicted onset of bulimic pathology ( Stice, Killen, Hayward, & Taylor, 1998a ), suggests that there may be reciprocal relations between these constructs. It is possible that there is a feedback loop operating, wherein adolescents begin to binge and purge in an effort to regulate their affect, which results in even greater affective disturbances, and so on. These results suggest that it may be fruitful for future studies to investigate the possible reciprocal relations between affective disturbances and bulimic pathology in greater detail. Limitations of the Current Study Although this study improved on past research by using multiple-method data collection, using structured interviews to assess major depression diagnoses, and following a large sample of youths over a 4-year period, the limitations of this investigation should be noted. First, this preliminary study was able to only partially test the gender-additive model of depression proposed here because we did not include boys in the sample. Second, these results are conditional in nature in that adolescents had to be depression free at study entry to be included in the analyses. The risk factors for early depression may be somewhat different than those for later depression. Third, the temporal reliability of our SCID major depression diagnoses (κ = .51) was only fair according to the criteria proposed by Fleiss (1981) , which likely resulted in a more conservative test of our hypotheses because this constrained our statistical power. Nonetheless, this test—retest reliability coefficient was somewhat higher than the average test— retest kappa of .46 for SCID I diagnoses from another community-based study ( Williams et al., 1992 ). Fourth, this study relied primarily on adolescent report data. Although adolescents are considered the most valid reporters for their own depressive symptoms ( Edelbrook, Costello, Dulcan, Kalas, & Conover, 1985 ), the fact that most of the data were self-report likely inflated the magnitude of the relations. Finally, the nonexperimental design limits the confidence that can be placed in the causal inferences. Randomized prevention trails that reduce these weight-related risk factors would be useful in


11/7/2001

Page 8 of 12

generating experimental evidence that these variables are causal risk factors for major depression. Research and Prevention Implications Collectively, these findings are consistent with the model suggesting that the body-image- and eatingrelated risk factors that emerge after puberty contribute to the elevated rates of depression among females. Because this is the first study to test whether this combination of risk factors prospectively predicts onset of interviewer-diagnosed major depression among female adolescents, the results have important research implications. First, although our findings suggest that these risk factors may partially explain the increases in depression among female adolescents, future studies will need to test whether these weight-related risk factors escalate in early adolescence for girls, but not boys, and whether the different levels of these risk factors account for the gender differences in depression. It is also likely that there are other important contributors to the emergence of gender differences in depression during adolescence (see Angold & Worthman, 1993 ; Nolen-Hoeksema & Girgus, 1994 ; Rutter, 1986 ). Second, future research might include other general risk factors for depression that appear to apply to both females and males (e.g., deficits in social support and negative life events; Lewinsohn et al., 1994 ) to more specifically test the hypothesis that the body-image- and eating-disturbance-related risk factors operate in addition to these more general risk factors. The fact that the risk factors examined here were significantly related to the onset of depression above and beyond the effects of negative affectivity (see footnote 2) provides preliminary evidence for this position. It would also be useful for future studies to begin at an earlier age so that participants can be tracked through the onset of puberty, which would permit a more explicit link between maturational changes and the emergence of these risk factors. Finally, because the dual pathway model posits that body-image and eating disturbances contribute to broad-based affective pathology, it is possible that these factors may also contribute to the development of anxiety symptoms. We hope that future research will explore this possibility as well. These findings also have prevention implications. This study identified several malleable risk factors that might be targeted in prevention efforts. Specifically, data suggest that programs that effectively reduce body dissatisfaction, dieting, and bulimic symptoms may prove useful in decreasing the incidence of major depression among female adolescents. These risk factors might also be used to identify groups who are at high risk for depression for targeted prevention efforts. Finally, it is also possible that future prevention efforts may be directed at both bulimic pathology and major depression, given the evidence that they may share some common risk factors. Ultimately, our understanding of the etiology of major depression might be advanced by a more thorough examination of the interrelations between mood problems and eating pathology.

References Allgood-Merten, B., Lewinsohn, P. M. & Hops, H. (1990). Sex differences and adolescent depression. Journal of Abnormal Psychology, 99, 55-63. American Psychiatric Association. (1987). Diagnostic and statistical manual for mental disorders ((3rd ed., rev.). Washington, DC: Author) Angold, A. & Worthman, C. W. (1993). Puberty onset of gender differences in rates of depression: A developmental, epidemiologic, and neuroendocrine perspective. Journal of Affective Disorders, 29, 145158. Baron, R. M. & Kenny, D. A. (1986). The moderator—mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182. Birmaher, B., Ryan, N., Williamson, D., Brent, D., Kaufman, J., Dahl, R., Perel, J. & Nelson, B. (1996).


11/7/2001

Page 9 of 12

Childhood and adolescent depression: A review of the past 10 years. Part I. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 1427-1439. Cole, D. A., Martin, J. M., Peeke, L. G., Seroczynski, A. D. & Hoffman, K. (1998). Are cognitive errors of underestimation predictive or reflective of depressive symptoms in children? A longitudinal study. Journal of Abnormal Psychology, 107, 481-496. Cox, D. R. (1970). Analyses of binary data. (London: Mathuen) Edelbrook, C., Costello, A. J., Dulcan, M. K., Kalas, R. & Conover, N. C. (1985). Age difference in the reliability of the psychiatric interview of the child. Child Development, 56, 265-275. Fabian, L. J. & Thompson, J. K. (1989). Body image and eating disturbance in young females. International Journal of Eating Disorders, 8, 63-74. Fleiss, J. L. (1981). Statistical methods for rates and proportions ((2nd ed.). New York: Wiley) Garner, D. M., Olmsted, M. P. & Polivy, J. (1983). Development and validation of a multidimensional eating disorder inventory for anorexia nervosa and bulimia. International Journal of Eating Disorders, 2, 15-34. Garrow, J. S. & Webster, J. (1985). Quetlet's index (W/H2) as a measure of fatness. International Journal of Obesity, 9, 147-153. Gotlib, I. H., Lewsinsohn, P. M. & Seeley, J. R. (1998). Consequences of depression during adolescence: Marital status and martial functioning in early adulthood. Journal of Abnormal Psychology, 107, 686-690. Graber, J. A., Brooks-Gunn, J., Paikoff, R. L. & Warren, M. P. (1994). Prediction of eating problems: An 8-year study of adolescent girls. Developmental Psychology, 30, 823-834. Hankin, B. L., Abramson, L. Y., Moffitt, T. E., Silva, P. A., McGee, R. & Angell, K. E. (1998). Development of depression from preadolescence to young adulthood: Emerging gender differences in a 10-year longitudinal study. Journal of Abnormal Psychology, 107, 128-140. Hayward, C., Killen, J. D., Kraemer, H. C. & Taylor, C. B. (1998). Linking childhood behavioral inhibition to adolescent social phobia: A prospective study. Journal of the American Academy of Child and Adolescent Psychiatry, 37, 1-9. Heatherton, T. F., Herman, C. P., Polivy, J., King, G. A. & McGree, S. T. (1988). The (mis) measurement of restraint: An analysis of conceptual and psychometric issues. Journal of Abnormal Psychology, 97, 19-28. Keel, P. K., Fulkerson, J. A. & Leon, G. R. (1997). Disordered eating precursors in pre- and early adolescent girls and boys. Journal of Youth and Adolescence, 26, 203-216. Killen, J. D., Taylor, C. B., Telch, M. J., Robinson, T. N., Maron, D. J. & Saylor, K. E. (1987). Depressive symptomology in adolescent binge eaters and purgers. American Journal of Public Health, 77, 1539-1541. Kraemer, H. C., Berkowitz, R. I. & Hammer, L. D. (1990). Methodological difficulties in studies of obesity I: Measurement issues. Annals of Behavioral Medicine, 12, 112-118. Lewinsohn, P. M., Gotlib, I. H. & Seeley, J. R. (1995). Adolescent psychopathology IV: Specificity of psychosocial risk factors for depression and substance abuse in older adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 34, 1221-1229. Lewinsohn, P. M., Hops, H., Roberts, R. E., Seeley, J. R. & Andrews, J. A. (1993). Adolescent psychopathology: I. Prevalence and incidence of depression and other DSM—III—R disorders in high school students. Journal of Abnormal Psychology, 102, 133-144. Lewinsohn, P. M., Roberts, R. E., Seeley, J. R., Rohde, P., Gotlib, I. H. & Hops, H. (1994). Adolescent psychopathology: II. Psychosocial risk factors for depression. Journal of Abnormal Psychology, 103, 302-315. Luke, D. A. (1993). Charting the process of change: A primer on survival analysis. American Journal of Community Psychology, 21, 203-246. Newcomb, M. D. & Bentler, P. M. (1988). Impact of adolescent drug use and social support on problems of young adults: A longitudinal study. Journal of Abnormal Psychology, 97, 64-75. Nolen-Hoeksema, S. (1994). An integrative model for the emergence of gender differences in depression


11/7/2001

Page 10 of 12

in adolescence. Journal of Research on Adolescence, 4, 519-534. Nolen-Hoeksema, S. & Girgus, J. S. (1994). The emergence of gender differences in depression during adolescence. Psychological Bulletin, 115, 424-443. Radloff, L. S. (1977). A CES—D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401. Rierdan, J., Koff, E. & Stubbs, M. L. (1989). A longitudinal analysis of body image as a predictor of the onset and persistence of adolescent girls' depression. Journal of Early Adolescence, 9, 454-466. Robert, R. E., Lewinsohn, P. M. & Seeley, J. R. (1991). Screening for adolescent depression: A comparison of depression scales. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 58-66. Rosen, J. C., Tacy, B. & Howell, D. (1990). Life stress, psychological symptoms and weight reducing behavior in adolescent girls: A prospective study. International Journal of Eating Disorders, 9, 17-26. Rosenberg, M. (1979). Conceiving the self. (New York: Basic Books) Rutter, M. (1986). The developmental psychopathology of depression: Issues and perspectives.(In M. Rutter, C. E. Izard, & P. B. Read (Eds.), Depression in young people: Developmental and clinical perspectives (pp. 3—30). New York: Guilford Press.) Segal, D. L., Hersen, M. & Van Hasselt, V. B. (1994). Reliability of the Structured Clinical Interview for DSM—III—R: An evaluative review. Comprehensive Psychiatry, 35, 316-327. Shisslak, C. M., Pazda, S. L. & Crago, M. (1990). Body weight and bulimia as descriptors of psychological characteristics among anorexic, bulimic, and obese women. Journal of Abnormal Psychology, 99, 380-384. Spitzer, R. L., Williams, J. B., Gibbon, M. & First, M. B. (1990). Structured Clinical Interview for DSM—III—R ((SCID). Washington, DC: American Psychiatric Press) Stice, E., Killen, J. D., Hayward, C. & Taylor, C. B. (1998a). Age of onset for binge eating and purging during adolescence: A four-year survival analysis. Journal of Abnormal Psychology, 107, 671-675. Stice, E., Killen, J. D., Hayward, C. & Taylor, C. B. (1998b). Support for the continuity hypothesis of bulimic pathology. Journal of Consulting and Clinical Psychology, 66, 784-790. Stice, E., Nemeroff, C. & Shaw, H. (1996). A test of the dual pathway model of bulimia nervosa: Evidence for restrained-eating and affect-regulation mechanisms. Journal of Social and Clinical Psychology, 15, 340-363. Strakowski, S. M., Tohen, M., Stoll, A. L., Faedda, G. L., Mayer, P. V., Kolbrener, M. L. & Goodwin, D. C. (1993). Comorbidity in psychosis at first hospitalization. American Journal of Psychiatry, 150, 752-757. Tabachnick, B. G. & Fidell, L. S. (1989). Using multivariate statistics ((2nd ed.). New York: HarperCollins) Tatsuoka, M. M. (1971). Multivariate analysis: Techniques for educational and psychological research. (New York: Wiley) Willett, J. B. & Singer, J. D. (1993). Investigating onset, cessation, relapse, and recovery: Why you should, and how you can, use discrete-time survival analysis to examine event occurrence. Journal of Consulting and Clinical Psychology, 61, 952-965. Williams, J. B., Gibbon, M., First, M., Spitzer, R. L., Davies, M., Borus, J., Howes, M., Kane, J., Pope, H., Rounsaville, B. & Wittchen, H. (1992). The Structured Clinical Interview for DSM—III—R (SCID): Multisite test—retest reliability. Archives of General Psychiatry, 49, 630-636. 1 Sexual maturation level was not included in the current analyses as a risk factor because nearly all of the female adolescents (92%) had already completed puberty by study entry.


11/7/2001

Page 11 of 12

2 Because a previous study from this project indicated that temperamental negative affectivity predicted onset of depression over the study period among both male and female participants ( Hayward et al., 1998 ), we tested whether the body-image- and eating-related risk factors examined here predicted onset of depression over and above negative affectivity in the female subsample. Post hoc analyses revealed that when initial negative affectivity was entered first in the models examining each risk factor separately, body dissatisfaction, dietary restraint, and bulimic symptoms still predicted onset of depression (all p s < .01). The pattern of findings was identical when initial depressive symptoms were entered into the equations as well. Thus, the bivariate relations between these risk factors and the onset of depression remained when negative affectivity was controlled statistically. 3 The evidence that self-esteem is correlated with body dissatisfaction, dieting, bulimic symptoms, and depression (e.g., Shisslak, Pazda, & Crago, 1990 ) raises the possibility that this third variable accounted for the prospective effects reported here. However, post hoc analyses revealed that self-esteem at study entry (as assessed by the Rosenberg Self-Esteem Scale; Rosenberg, 1979 ) was not significantly related to the hazard for onset of major depression ( B = 0.43, hazard ratio = 1.54, ns ). Lewinsohn, Gotlib, and Seeley (1995) similarly found that initial self-esteem did not predict onset of major depression during adolescence. These findings collectively suggest that initial deficits in self-esteem cannot account for the observed relations. 4 Although discriminant function analyses do not permit the use of covariates, we estimated this model including initial depressive symptoms, but this variable did not improve the model fit, χ 2 (4, N = 1120) = 25.63, p < .0001, or alter the correct classification rate.

This study was supported by National Institute of Mental Health Postdoctoral Fellowship MH19908, Career Award MH01708, and Grant MH45431; National Institute for Child Health and Development Grants HD24240 and HD24779; Stanford Center on Adolescence; and the W. T. Grant Foundation Faculty Scholar Award. The manuscript was completed when Eric Stice was a visiting assistant professor of psychiatry at Stanford University School of Medicine. Thanks go to Helena Kraemer and Anne Varady for their thoughtful comments regarding drafts of this article. We thank Anne Blair-Greiner, Darby Cunning, Diane Strachowski, Suzanne Taborski-Barba, and the Santa Clara Unified and Fremont Union School Districts for their help in conducting this study. Correspondence may be addressed to Eric Stice, Department of Psychology, University of Texas at Austin, 330 Mezes Hall, Austin, Texas, 78712. Electronic mail may be sent to [email protected] Received: June 7, 1999 Revised: November 1, 1999 Accepted: December 15, 1999

Table 1. Means, Standard Deviations, and Bivariate Correlations Among Predictors at Time 1


11/7/2001

Page 12 of 12

Table 2. Unstandardized Regression Coefficients, Hazard Ratios, and 95% Confidence Intervals From the Hazard Analyses Predicting Onset of Major Depression


11/7/2001